CC* Compute: Building a state-of-the-art campus compute resource at Franklin & Marshall College

点击翻译按钮获取中文摘要

Franklin & Marshall College (F&M) is building and deploying a campus cluster resource to better meet the needs of our researchers and their students who need greater access to high performance compute resources to support intensive data analysis and computation. This project provides much needed local compute nodes for F&M's researchers and students while also contributing to the growing fabric of shared computing clusters across the country. This project contributes these new resources to the Open Science Grid (OSG) which is a national, distributed computing partnership that allows participants to share their resources with other researchers to maximize the impact these investments have on scientific research and discovery. As an institution, F&M has many top-tier scientific researchers who also partner with students in research that is regularly funded by public and private agencies. Providing substantially improved infrastructure to support this research will advance and expand the institution's capacity to support important investigations, from the search for pulsars to brain science. F&M has a demonstrated commitment to recruiting and supporting STEM students and this infrastructure improvement and investment allows the institution to continue to be a leader in this arena, providing access to the best resources and opportunities for future scientists. This project, similar to other recent initiatives, demonstrates how it is possible to design and implement significant research infrastructure, even at a smaller institution, that advances scientific discovery both on and beyond our campus.The compute cluster maximizes available resources for research that requires both HPC and HTC solutions. It consists of a head node and a login node, both running dual 20 core Xeon Gold 6230 2.10GHz CPUs with 192GB of 2666 MHz ECC memory, and 38TB of SSD NVMe storage. There are 36 standard compute nodes, each using dual 20 core Xeon Gold 6230 2.10 GHz CPUs with 192GB of 2666 MHz ECC memory and 1.8TB NVMe SSD for OS and local scratch. There is one GPU node for use with software that takes advantage of cuda compiled software and GPU co-processing. It mirrors the same specs as the standard compute node, but includes four Nvidia V100 GPU cards featuring 32 GB HBM2 memory and 5120 stream processors.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.